煤矿新生采空区空洞流算法研究

The newborn goaf in the range of step distance of main roof caving behind working face is a sensitive area, which affects greatly spontaneous combustion and concentration distribution. In this area there are many cavities that are probably a few meters or even tens of meters, and it results in deviation from the hypothesis of REV in porous media hydrodynamic. That's why there is a greater error between traditional CFD simulation results and the scene situation. The near-trapezoidal stage newborn goaf formed by single mining face is taken as the object of study, which is described by binary coding based on three-dimensional lattice lattice. Genetic algorithm is adopted for solving the fast convergence to real falling form and the real gas flow, and the gas flow and the gas dispersion state of the cavity flow in the newborn goaf is solved by The lattice boltzmann algorithm. The system error between the results of the solution and measured data of the boundary or internal measurement points of the real goaf (experimental model) is used as the fitness function.In order to solve the problem of mass operation, Bayesian network is used to improve the initial population sample generation, crossover and mutation operator algorithms by using Monte Carlo method and geometric fractal statistics rules of goaf. The finite model samples are established by using the physical experiment model, and finite flow samples are gotten by changing the boundary conditions, and it is used to verify the feasibility and accuracy of the method.

采煤工作面后方老顶垮落步距尺度范围内的采空区（新生采空区）空洞空间往往达几米甚至几十米，偏离多孔介质流体动力学表征体元REV的假设前提,这是基于多孔介质流体动力学的传统CFD模拟结果与现场实际情况有较大误差的根本原因。项目以单一回采工作面形成的近梯形台型新生采空区为研究对象，用三维网格点阵二进制编码描述采空区，用遗传算法解决“穷举采空区”快速向真实冒落形态与真实气相流态收敛的问题。用格子Boltzmann法解算新生采空区空洞流气体流动状态和组分气体弥散状态，将解算结果与真实采空区（实验模型）的边界或内部测点实测数据进行比较，把系统误差作为适应函数。为解决海量运算问题，综合采用蒙特卡罗法以及用于描述采空区冒落几何分形统计规则的贝叶斯网络对初始种群样本生成、交叉和变异算子算法进行改进。用物理实验模型模拟典型冒落样本，改变物理模型边界条件模拟典型流态样本，以此验证本方法的可行性与准确性。

准确描述采空区气体流动状态是防治遗煤自燃和瓦斯积聚的基础。课题研究发现，以往建立在REV尺度上的采空区渗流模拟，假设新生采空区为连续介质，忽略冒落形态进行计算。无法真实的反映出气体在采空区复杂几何结构内部的流动细节，计算结果与现场实际有较大的差异。通过研究发现，新生采空区内部的冒落岩石块度大，岩块与岩块之间，岩块与顶板间都有较大的空洞，气体在这些空洞间流动遵循低粘度不可压缩流动。多孔介质流体动力学不适用于新生采空区。本课题为了准确描述新生采空区气体流动状态，提出了新生采空区空洞流算法，利用遗传算法（GA）随机生成采空区冒落形态二值矩阵，将格子Boltzmann方法（LBM）解算的气体流动状态与真实采空区的边界或内部测点实测数据误差作为GA适应值，通过GA快速逼近真实采空区冒落形态和气体流动状态。为了验证算法有效性，搭建了新生采空区实验模型，进行了激光多普勒测速（LDA）实验。结果表明，算法能够通过采空区内部少量观测点数据反演出采空区的冒落形态和气体流动状态，并且反演的冒落形态和气体流动状态与实际情况基本一致。实际与重构冒落形态的杰卡德相似度达到了0.7473，解算得到的风速值与LDA实测值误差较小，均方误差为0.0244，R2系数达到了0.8986。新生采空区空洞流算法解决了多孔介质流体动力学不适用于新生采空区的问题。解决了采空区冒落形态难以探测和重构的问题，为采空区气体流动状态解算提供了一种新方法。